I'm sitting at my computer, playing a game. It's not a typical game: I'm using human spatial reasoning and puzzle-solving know-how to manipulate and shape virtual proteins.
The game—FoldIt—is an exercise in molecular origami. I use my mouse to tug and twist at a backbone of mottled greens, browns, oranges and reds on my screen, each color representing the properties of a particular region of the protein. Side chains, chemical pendants that make the protein’s building blocks unique, hang off the main backbone like charms on a bracelet.
Proteins are long chains of building blocks called amino acids, the specific number and arrangement of which makes each protein—whether it makes up your hair or carries oxygen in your blood—unique. In the cell, proteins fold as they are assembled, the chain (or backbone) of molecules twisting and kinking to make a structure that resembles a tangled Slinky. A protein's shape (or structure) determines what it does, where it goes, and the molecules with which it interacts.
At the moment I'm working on a poisonous protein produced by the funnel spider. The protein is clearly unhappy: unnaturally elongated, its color palate is more angry red than green, and four atoms have been flagged as too close to one another for comfort.
A few simple moves yield big dividends: From a starting score of 1,807, two quick keystrokes make the protein noticeably more compact. The atoms get their space, and the color palate has shifted toward green. My score now at 7,710, the current high score—8,649—seems within reach. Yet I'm at a loss of how to get there.
I have a PhD in cell and molecular biology from the University of Pennsylvania, but it's not enough. Frustrated, and with my player ranking at a dismal 430th out of 450, I give up.
Having a doctorate means I know how laborious and expensive it is to determine the correct structure for a given protein in the lab. A relatively short protein of, say 100 amino acids, could assume trillions of different shapes. Only one is correct—typically the one with the lowest energy. That's because, as University of Washington (U.W.) in Seattle biochemist David Baker explains it, a protein's structure is like a ball on a sloping floor: It will find its lowest energy state just as the ball will naturally roll to the surface's lowest point. Figuring out the "correct" shape of a given protein, then means finding the shape with the lowest energy level.
Baker came up with an automated way to do that: Rosetta@home. Like the popular SETI@home screen saver that is used to help sift out any signal from the cosmos that may be of intelligent origin, Rosetta harnesses processing power from idle computers around the world to predict protein shapes, twisting and bending chains to try to get to the minimum energy. Sometimes, the program makes rookie mistakes. Users saw them: "I'm watching what's going on on my computer, and these random moves the computer's making," Baker recalls hearing, "are often just silly."
A colleague, David Salesin, suggested converting Rosetta@home into an interactive game. He connected Baker with Zoran Popović, a computer scientist at U.W., who in turn passed the project to his graduate student, Seth Cooper, and postdoc Adrien Treuille. The first public beta was unveiled a year later.
Your challenge if you download the software: to pull, push, nudge and rotate the protein, represented as a three-dimensional, multicolored pipe, into its correct shape using tools such as pull and tweak, shake and wiggle. Each structure is assigned a score: the lower the energy level, the higher the score. Introductory exercises and in-game aides like "peekaboo," which compares top-scoring solutions with yours, help novices get up to speed.
For me, there are too many options. For top-10 player (and former number one) Sirenbrian—aka Florida-based software engineer Brian Smith, who plays FoldIt a couple of hours every night—the signature move is known as the "local wiggle strategy" as well as "walking the backbone," "slice 'n dice" or just "Brianizing". He isolates short segments of the structure and "wiggles" to lower the energy of the locked-off segment.
Former number one player Charles Cusack (screen name "Ferzle"; he's now ranked 23rd), an assistant professor of computer science at Hope College in Holland, Mich., uses a very simple strategy: pull on the protein's backbone, shake, wiggle and repeat. "You just change it a little bit," he says "then let the algorithms do their work."
Staying up late playing protein-folding games may seem a lonely exercise, but chat windows, a wiki, duels and group play make FoldIt into a social environment in which users learn from each other. More than 50,000 individuals have downloaded the application since its release in May. Hundreds actively play it, both alone and in groups, with five or six new gamers joining hourly. Only about half are biologists; the others range from software engineers to historians, from grandmothers to middle-schoolers.
There aren't any prizes, although Baker acknowledges the winners on the FoldIt Web site. Yet for users, it's about more than just the high score. "It is a game that feeds into the actual scientific process," Sirenbrian says. "We might be in the process of developing a new way of analyzing proteins and being of help to people." Thrianya, a grandmother of three with a high school education, says in her user profile, "I love pushing these little bits and pieces around and enjoy very much chatting with people from around the world who like the same sort of thing."
Baker and his team learn from their users, continually on the lookout for moves that users rely on. The most successful routine manipulations will ultimately be codified into Rosetta@home, to make the screen saver more effective.
FoldIt's developers also fly top folders to Seattle to watch them work with the program. Sirenbrian spent a day and a half at FoldIt headquarters in early August. "They interviewed me, videotaped me playing a little bit, and we talked about some of the new tools and features I would like to see added," he says. Among his suggestions were a "squeeze" function, to compress the entire structure (more compact proteins tend to have lower energy) and the ability to control the strength of the wiggle function.
One of Sirenbrian's biggest complaints with FoldIt is that the game is short on instructions. "It is like throwing a lot of tools into the pit," he says, "and seeing what we do with them."
From Baker's perspective, however, that is precisely the point. "We don't know a priori what the best strategies are [for folding proteins], so we create the tools and see what people do with them."
Still, U.W.'s Popović says user input has doubled the number of FoldIt tools since its initial release in May. There's now an annotation tool, which allows users to tag a structure with notes for others to view as well as a tool that rotates structural elements.
Soon, the developers will rely on their users' wisdom to design totally new proteins, which FoldIt has planned for the next release, scheduled to go live sometime in October or November. The new version challenges users to design completely novel proteins that could become tomorrow's HIV vaccines and biofuels. Baker plans to create the 10 best solutions for each challenge in his lab to see how they work.
"We are hoping to define this whole new genre of games, that we would like to call 'scientific discovery games,'" Popović says. "The idea is to find many different places in science where the human ability to problem solve can be directly applied without necessarily requiring somebody to get a PhD in a particular field first."